# Inkling Small

Inkling Small is about a quarter the size of Inkling and keeps most of what makes the family useful: native reasoning over text, images, and audio, controllable thinking effort, and strong agentic coding and tool use.

- **Model ID:** `thinkingmachines/inkling-small`
- **Type:** chat
- **Providers:** deepinfra, thinkingmachines
- **Context window:** 1,000,000
- **Maximum output tokens:** 1,000,000
- **Pricing:** $0.3/1M input tokens, $1.2/1M output tokens
- **Canonical page:** https://vercel.com/ai-gateway/models/inkling-small

## Supported parameters

Detailed capability metadata has not been reported for this model.

## Example

```ts
import { streamText } from 'ai'

const result = streamText({
  model: 'thinkingmachines/inkling-small',
  prompt: 'Why is the sky blue?'
})
```

## About

Inkling Small became available on AI Gateway on July 30, 2026. Inkling Small is a 42-layer decoder-only transformer with a sparse Mixture-of-Experts (MoE) backbone: 276 billion total parameters, 12 billion active per token, and each token routed to 6 of 256 experts plus 2 shared experts. Attention mixes local and global layers, images enter through a hierarchical patch encoder, audio through discrete token encoding, and the context window is 1M tokens. Thinking Machines released the weights under the Apache 2.0 license.

On agentic coding, Inkling Small scores 80.2% on SWE-bench Verified, 55.9% on SWE-bench Pro (public), 64.7% on Terminal-Bench 2.1, and 48.7% on SciCode. Each of those sits at or above Inkling, which scores 77.6%, 54.3%, 63.8%, and 46.1%. Tool use follows the same pattern: 54.4% on Toolathlon Verified against Inkling's 45.5%, and 79.6% on the public split of MCP Atlas. On reasoning it scores 89.5% on GPQA Diamond, 90.2% on HMMT February 2026, 31.6% on Humanity's Last Exam text-only, and 40.1% on ARC-AGI-2.

The gap is knowledge. Inkling Small scores 20.6% on SimpleQA Verified where Inkling scores 43.9%, and it trails on the AA Omniscience index and on Tau 3 Banking, a multi-turn domain agent evaluation. Audio results sit close behind Inkling at 90.1% on VoiceBench, 77.0% on MMAU, and 54.9% on Audio MC. Treat Inkling Small as a capable reasoner with a smaller store of memorized facts, and give it a retrieval path when questions turn factual.

Vision is a practical strength. Inkling Small scores 74.0% on MMMU Pro and 77.4% on CharXiv reasoning questions, rising to 81.3% when it crops, zooms, and inspects images programmatically. That helps most on documents, forms, and charts where the detail that answers the question is small.

Controllable thinking effort runs from minimal to maximum, so you can trade answer quality against cost and latency per request. Sweep the setting across representative traffic rather than fixing it up front.

Inkling Small is compatible with Zero Data Retention on AI Gateway. Turn it on team-wide from the dashboard, or per request with `zeroDataRetention: true`, and AI Gateway routes only to providers that delete prompts and responses after each request. Set the model to `thinkingmachines/inkling-small` in the AI SDK, Chat Completions API, Responses API, Messages API, or other API formats, from TypeScript or Python. To use Inkling Small in a coding agent, run `vercel ai-gateway setup`, then select `thinkingmachines/inkling-small` in the agent's model configuration.

## What to consider

Size shows up in world knowledge, not in reasoning or coding. Inkling Small scores 20.6% on SimpleQA Verified against Inkling's 43.9%, so anything that depends on recalling specific facts needs retrieval or web search alongside the model. Reasoning, coding, and tool-use scores hold up, and several of them land above Inkling's.

## When to use

### Best For

- **High-Volume Coding Agents:** Agentic coding and tool use, with 80.2% on SWE-bench Verified
- **Tool Orchestration Pipelines:** 54.4% on Toolathlon Verified and 79.6% on the public MCP Atlas split
- **Document and Chart Analysis:** Programmatic cropping and zooming to read small visual detail
- **Compact Multimodal Apps:** Native text, image, and audio input in a smaller model than Inkling
- **Effort-Tuned Latency Budgets:** Thinking effort set from minimal to maximum on each request
- **Zero Data Retention Routing:** Requests routed only to providers that delete prompts and responses

### Consider Alternatives When

- **World Knowledge Questions:** 20.6% on SimpleQA Verified against Inkling's 43.9% is a real recall gap
- **Strongest Audio Results:** Inkling scores higher on VoiceBench, MMAU, and Audio MC
- **Multi-Turn Domain Agents:** Inkling scores higher on Tau 3 Banking and on broader factual evaluations
- **Frontier Coding Ceiling:** Closed frontier models still lead on SWE-bench Verified and Terminal-Bench 2.1

## Best for

- **High-Volume Coding Agents:** Agentic coding and tool use, with 80.2% on SWE-bench Verified
- **Tool Orchestration Pipelines:** 54.4% on Toolathlon Verified and 79.6% on the public MCP Atlas split
- **Document and Chart Analysis:** Programmatic cropping and zooming to read small visual detail
- **Compact Multimodal Apps:** Native text, image, and audio input in a smaller model than Inkling
- **Effort-Tuned Latency Budgets:** Thinking effort set from minimal to maximum on each request
- **Zero Data Retention Routing:** Requests routed only to providers that delete prompts and responses

## Consider alternatives

- **World Knowledge Questions:** 20.6% on SimpleQA Verified against Inkling's 43.9% is a real recall gap
- **Strongest Audio Results:** Inkling scores higher on VoiceBench, MMAU, and Audio MC
- **Multi-Turn Domain Agents:** Inkling scores higher on Tau 3 Banking and on broader factual evaluations
- **Frontier Coding Ceiling:** Closed frontier models still lead on SWE-bench Verified and Terminal-Bench 2.1

## Frequently asked questions

### How does Inkling Small differ from Inkling?

Size and knowledge. Inkling Small has 276 billion total parameters and 12 billion active, against 975 billion and 41 billion for Inkling. Inkling Small matches or beats Inkling on coding, reasoning, and tool-use evaluations, and trails it on factual recall, scoring 20.6% on SimpleQA Verified against 43.9%.

### How did Inkling Small score on coding benchmarks?

80.2% on SWE-bench Verified, 55.9% on SWE-bench Pro (public), 64.7% on Terminal-Bench 2.1, and 48.7% on SciCode. Each result sits at or above Inkling on the same evaluation.

### What input types does Inkling Small accept?

Text, images, and audio, with text output. Images are encoded through a hierarchical patch encoder and audio through discrete token encoding, and the decoder processes all three together.

### How does Inkling Small handle documents and charts?

Inkling Small can crop, zoom, and inspect images programmatically, which helps when the detail that answers a question is small. Inkling Small scores 77.4% on CharXiv reasoning questions, or 81.3% with a Python tool, and 74.0% on MMMU Pro.

### How does controllable thinking effort work?

You set effort per request, from minimal to maximum. Higher effort spends more thinking tokens and raises scores on hard tasks; lower effort cuts cost and latency. Published benchmark results are reported at maximum effort.

### Can I use Inkling Small in a coding agent?

Yes. Run `vercel ai-gateway setup` to connect your agents to AI Gateway, then select `thinkingmachines/inkling-small` in the agent's model configuration. See the coding agents guide at https://vercel.com/docs/ai-gateway/coding-agents.

### Does AI Gateway support Zero Data Retention for Inkling Small?

Yes, Zero Data Retention is available for this model. Turn it on team-wide from the dashboard or per request with `zeroDataRetention: true`. Zero Data Retention is offered on a per-provider basis. See https://vercel.com/docs/ai-gateway/capabilities/zdr for details.

### Are the weights for Inkling Small open?

Yes. Thinking Machines released Inkling Small with open weights under the Apache 2.0 license, which permits commercial use and modification.

### How do I call Inkling Small through AI Gateway?

Set the model to `thinkingmachines/inkling-small` in the AI SDK, Chat Completions API, Responses API, Messages API, or other API formats, from TypeScript or Python. AI Gateway handles authentication, retries, and failover across `deepinfra`, `thinkingmachines`.

### What does Inkling Small cost?

Current rates appear in the pricing panel on this page. AI Gateway mirrors provider pricing with no markup and adds no platform fee on inference, including on Bring Your Own Key requests.

## Links

- [Model page](https://vercel.com/ai-gateway/models/inkling-small)
- [AI Gateway documentation](https://vercel.com/docs/ai-gateway)
- [Provider model documentation](https://thinkingmachines.ai/news/introducing-inkling/)